# Full text of "The Husimi function and a successive measurement of the position and the momentum"

## See other formats

```The Husimi function and a successive measurement of the
position and the momentum

Takuro Shitq. s
University of Electro- communications, Chofu, Tokyo 182-8585, Japan

Abstract

In this paper, we consider an interpretation of the Husimi function as the probability distribution
of a successive measurement, which is clearly separated into measurements of the position and the
momentum. We also show this successive measurement can be easily realized in the situation close
to the experiment, and this measurement corresponds to one of Aharanov's weak measurement.

* |sito taku@ quest.i s. uec. ac.jp

1

I. INTRODUCTION

In this paper, we will discuss properties and structures of the Husimi function as the
probability distribution of a successive measurement of the position and the momentum.

In the literature, Wigner et al. discussed the distribution of the position and the mo-
mentum, defined so called the Wigner function j^J, and considered its properties in 1930s,
while Husimi et al. discussed the distribution of the position and the momentum, and de-
fined so called Husimi function [1]. In the Wigner function, there is a problem that the
Winger function is not limited to a positive function. Feynman similarly discussed the neg-
ative probabilities at the spin system Holevo discussed a simultaneous measurement of
the position and the momentum in 1970s, he showed a measurement process of the Husimi
function by using Naimark extension jg]. Recently Aharanov et al. (4] proposed a weak
measurement and a weak value which is similar to the negative probability.

In our study, we discussed the meaning of the marginal distribution of the Husimi func-
tion, and discovered that the Husimi function can be obtained by Aharanov's weak measure-
ment. First, we show that the measurement to obtain the Husimi function can be clearly
separated into two measurements of the position and the momentum. Secondly, we show that
these measurements can be easily realized in the situation close to the experiment. Finally,
we discuss the relation between these measurements and Aharanov's weak measurement.

II. DEFINITION AND REVIEW OF THE HUSIMI FUNCTION

The Wigner function, the Husimi (Q) function, and the Glauber-Sudarshan (P) function
are known as distributions including complete information of the quantum state. In this
section, we will define and review the Husimi function and other distribution functions jj— 13 ] .
These distribution functions are defined by the following family of characteristic functions
parameterized by a parameter s:

w(u, v, s) = T T p e -™z-ivp e i(u +v ) ?

2

where p is the density operator of a state satisfying p > and Trp = 1. Using these
characteristic functions, we define

W ? (x,p) = J I dudvw(u, v, )e iM+ ^, (2)

Qp{x,p) = J j dudvw(u,v, -i) e ^+^, (3)

Pp(x, p) = j±y 2 JJ dudvw(u, v, +l) e ^+^, (4)

where W p (x,p) is the Wigner function (s = 0), Qp(x,p) is the Husimi function (s = —1),
and P p (x,p) is the Glauber-Sudarshan function (s = 1). Here and hereafter the range of the
integral is from — oo to oo and is omitted.

While these functions are defined for density operators, we can also define them with
respect to any linear operators by linearity. Moreover the trace of the product of any two
linear operators can be written by Wigner functions.

Tr AB = JJ dxdpW A (x,p)Wz(x,p). (5)

Using this fact, the expectation of any observable A for a state p is written by Wigner
functions.

(A) p = TrpA= JJ dxdpW p (x,p)W A (x,p). (6)

Similarly, using the Husimi function and the Glauber-Sudarshan function, we can also write
the expectation of an observable A as follows:

(A) p = TrpA = JJ dxdpQ p (x,p)P A (x,p) (7)

= JJ dxdpP p (x,p)Q A (x,p). (8)

We can derive a simple formula of the Wigner function and the Husimi function for a
state p respectively,

W p (x,p) = ^-J dye™ (x-y/2\p\x + y/2) , (9)

Qp( x ip) = 2~ ( x >p\p\ x >p) > ( 10 )

where \x,p) is a "coherent state," i.e., an eigenvector of the annihilation operator a. It can
be seen from ( fit)]) that the Husimi function is always positive, while the Wigner function and
the Glauber-Sudarshan function sometimes take negative values. The marginal distribution

of the position x for the Husimi function and the Wigner function is, respectively, written
by

dpWp(x,p) = (x\ p \x) , (11)
J dpQ p (x,p) = -1= J dx'e- {x - x ' ? (x'\ p \x') . (12)

In the subsequent discussion, we consider properties of the Husimi function. The equation
( flOl) is frequently used as a definition of the Husimi function. In our discussion we follow
this definition, because this definition can include a freedom of the squeezed parameter A as
follows. A squeezed coherent state is expressed by the vacuum state |0), which is a minimum
uncertainty state, i.e., the state satisfying AxAp = 1/4. Let 5(A) be the squeezed operator
and D(x,p) be the shift operator. Then a squeezed coherent state is expressed by

\x,p;A) = D(x,p)S(A)\0)

= T^yf 4 J dx'e-^ x '- x ^ +lx ' p \x') . (13)

Now the Husimi function based on the squeezed coherent state is defined by

Q{x,p; A) = — (x,p; A\p\x,p; A) . (14)
Ztt

When the squeezed parameter A = 1, this definition corresponds to the definition by the
characteristic function ([3]). In the following, we omit the squeezed parameter A in \x,p;A)
for simplicity.

III. HUSIMI DISTRIBUTION FUNCTION AND POST SELECTION

In this section we study the detail of a physical interpretation of the Husimi function.
Considering a pure state p = in the definition of the Husimi function ( 1T4"1) . we have

Q(x,p; A) = (x,p; A\ip) | 2 . The term (x,p; A\ip) can be calculated as follows:

(x,p;A\i>) = J dx'e-^ x ~ x 'y 2 -^(x'^) (15)

= V^(p\ Jx ^e-^ x -^W, (16)

where we used the spectral decomposition of x:

x = [ dx'x'\x')(x'\. (17)

Now let us define an operator M(x) = ^ 1/4 e 3s( x x ^ , then fll6p is rewritten as

(x,p\ip) = V2^(p\M(x) . (18)

In the following, we will interpret the square of the absolute value of the above equation as
the probability of a successive measurement of the position and the momentum. We will
also discuss a physical realization of the successive measurement in the next section. The
operator M(x) is a measurement operator, which fulfill flowing properties;

dxM 2 (x) = /, (19)

and the operator M(x) mean the reduction of the wave packet, thus the state after the
measurement is;

|V') ex M(x) |V) = J dx^(x')e-^-^ 2 \x') , (20)

where ip(x) = (x\i/j). This operator M(x) have other properties described below. The
coherent state \x,p) is generated by operating M(x) to some eigenstates of the momentum

|P>-

y/2wM{x) \p) = \x,p) (21)

When we measure the eigenstate of the momentum p with M(x), the state will become
the coherent state after the reduction of the wave packet. The expectation of the squared
operator M(x) is the marginal distribution of the Husimi function.

(VI M 2 (x) |V) = (VI -^e-^ ix - £)2 |V)
V An

= f dx'e-^ x ' x ' )2 \ (x'|V) I 2 (22)

VA7T J

d P Q(x,p) (23)

However, this distribution does not have complete information of the state |V)- The operator
M(x) is written by a coherent state \x,p) as follows:

M(x) = | J dp\x,p) (x,p\ J . (24)

After we obtain the measurement result x of M(x), we have one to one correspondence
between the the state after the measurement |V') and the initial state ketip. In this sense,
the state after measurement |V') does not lose any information of the initial state |V)-

IV. PHYSICAL REALIZATION OF THE MEASUREMENT OF THE POSITION
AND POST SELECTION OF THE MOMENTUM

In this section, we consider a physical realization of the measurement M(x) in (j24j) . In
the following, we show that the Husimi function is obtained as a probability distribution of a
successive measurement, i.e., a weak measurement of the position followed by the projective
measurement of the momentum. Note that these measurements of the position and the
momentum are quite different in the sharpness, while the Husimi function as the probability
distribution has symmetry in the position and the momentum. It is known [6| that a
simultaneous measurement of the position and the momentum is possible on an enlarged
Hilbert space by Naimark extension. We can easily see that the measurement devised by
Holevo, is corresponding to one of measurements which obtain the Husimi function.

We observe some properties of the position measurement (|24p related to the squeezed
parameter A in (fl3l) . First, when A — > 0, i.e., the position measurement is sharp, following
probability distribution p(x) is obtained for any state

p(x) = hm (ip\M 2 (x) \if>)

lim [ dx'^=e-^ x - x ' )2 \ (M)

dx'5(x — x')\ (x'\i[)) |

= I W) I 2 - (25)

After this measurement, the initial state \ip) will be the eigenstate of the position \x), which
corresponds with obtained value x.

Secondly, when A — > oo, that is the unsharp limit of the position measurement, (121
corresponds with the identity operator for any x

lim Mix) = lim , A \ - e'^^'^ 2
a^oo v ' a^oo (Avr) 1 ^

oc

A

lim / dx'e ^ x \x') (x'\
dx' \x') (x'\

= I (26)

Thus we can not obtain any information of the system by the unsharp measurement
liniA-s.oo M(x). If we measure the momentum after the position measurement limA-^oo M(x),

the probability distribution is obtained as follows:

| (p\ lim M(x) |V) | 2 cx | (p\ I |V) | 2 = | (p|V) | 2 . (27)

A— >oo

We show that the measurement M(x) can be realized by a von Neumann measurement as
follows. We consider a composite system of the measurement device with squeezed vacuum
state |0; A) and the measured system with an initial state |V);

|0; A) |V) = J dx'e-^ x ' 2 \x') |V) , (28)

and the time evolution by the following interaction Hamiltonian;

Hinter = S(t - t')x^p md , (29)

where x^ operate on the measured system, p md operate on the system of the measurement
device and x^p m d means x^ ®p m d- The position of the measurement device is shifted by this
interaction Hamiltonian aXt — t' in the proportion to the position of the measured system.
The unitary operator of the time evolution between t = and t = t' is written as

U = exp | J dtH inter ^ = exp{-ix^p md }. (30)
After the interaction (!30|) . the composite system (!28l) will be written as

We measure the position of the measurement device and consecutively measure the momen-
tum of the measured system, then the probability of the position and the momentum is
given by

dx"e-^ x - x " )2 - ix " p (x"m , (32)

(p\ (x\ J dx'e-^ x - x ^ \x') |V)

which corresponds to the Husimi function of the measured system ( JT5|) .

In the following we show that this measurement and Aharanov's weak measurement
are related. Let us introduce a Hamiltonian corresponding to (|29|) by

4]

Hinter = ~gS(t - t')x^p md . (33)

Note that the above Hamiltonian includes the parameter g which means the weakness of
the interaction. We also consider the composite system corresponding to (1281) . without
parameter A:

|0) |V) = f dx'e- 1 ^' 2 \x') |V) ■ (34)

After the interaction by the Hamiltonian (133]) . the composite system AMI) will be

Then we measure the position of the measurement device and the momentum of the measured
system.

J dx"e-^ x - 9X " )2 - ix " p (x"\ip) = J dx"e~^ {x/9 - x " )2 - ix " p (x"\ip) (36)
Now we assume that A = 1/g 2 and x/g is x, then we obtain

dx"e-^ (x - x " )2 - ix " p (x"\ip) . (37)

This probability distribution corresponds with f[32|) . The Husimi function defined in (p2
includes the controllable parameter A, which means the sharpness of the position measure-
ment, while Aharanov et al. supposed that the interaction between the measured system
and the measurement device is so weak.

The momentum measurement in our measurement is the projective measurement, and is
clearly separated from the position measurement. Thus we can easily see that the conditional
probability distribution of the position, when the measured value of the momentum is p, is
given by

= «b (38)

where Q(xjj) is the Husimi function.

Holevo |6| already discussed a measurement to obtain the Husimi function by Naimark
extension. However, our measurement is different form Holevo's measurement in the point
that Holevo's measurement is simultaneous measurement in the extended Hilbert space by
Naimark extension. Additionally, in our measurements of the position and the momentum,
each measurements are clearly separated. On the other hand Holevo's measurements are not
separated, because the measurement of the relative position and the center momentum across
the extend Hilbert space are considered. Using our measurement, we can give comprehensible
physical meaning to equation fl38l) .

V. CONCLUSION

We proposed a new interpretation of the Husimi function including post-selection study-
ing the marginal distribution of the Husimi function and Aharanov's weak measurement.

S

We also showed that the probability distribution of our successive measurement of the po-
sition and the momentum corresponds to the Husimi function. This measurement process
is different from Holevo's measurement in the point that his measurement is a simultane-
ous measurement across the extended Hilbert space, while our measurement is a separated
successive measurement. The important point of our study is that our measurement can
be realized by the reduction of the wave packet of the measured system without consid-
ering the measurement device system. In the other words, our each measurements, the
position and the momentum, are separated. By using this fact, we can easily calculate the
marginal distribution or the conditional probability form the Husimi function like ( 1231) and
(I38p . Although our measurement is the simple combination of the weak measurement of
the position and the projective measurement of the momentum, the probability distribution
automatically corresponds to the Husimi function by the reduction of the wave packet. We
have also shown that our measurement process to obtain the Husimi function is the weak
measurement of the position with post-selection of the momentum.

ACKNOWLEDGEMENT

[1] K. Husimi, Proc. Phys. Soc. Jpn., 22, 264, 1940.

[2] E. P. Wigner, Phys. Rev., 40, 749, 1932.

[3] R. J. Glauber, Phys. Rev., 131, 2766, 1963.

[4] Y. Aharonov, D. Z. Albert, L. Vaidman Phys. Rev. Lett, 60, 1351, 1988.

[5] U. Leonhardt, Measuring the Quantum State of Light, Cambridge University Press, 1997.

[6] A. Holevo Probabilistic and Statistical Aspects of Quantum Theory, North Holland, 1982.

[7] M. A. Nielsen, I. L. Chuang, Quantum Computation and Quantum Information, Cambridge

University Press, 2000

[8] R. P. Feynman, Statistical Mechanics A Set of Lectures, Westview Press, 1998

9

```